BACKGROUND: Transbronchial lung biopsy (TBLB) has known to yield useful information for pulmonary infiltrates of uncertain etiology. However, the its safety and usefulness of TBLB has have not been conclusive in the critically ill patients with respiratory failure. Moreover, TBLB has not been recommended for patients with mechanical ventilation. This study was done conducted to investigate the diagnostic values and risks of TBLB performed on critically ill patients at bedside to obtain information on the pulmonary infiltrate of unknown etiology. METHODS: Twenty patients (21 admissions with 23 cases) with diffuse pulmonary infiltrates who were treated in a medical intensive care unit of a tertiary referral hospital from January 1994 to May 1998, were enrolled for this study. Their medical records were retrospectively reviewed. TBLB was opted when a noninvasive diagnostic work-up failed to reveal the cause for the pulmonary infiltrate. The procedure was performed at patients' bedside without assistance of fluoroscopy. Bronchial washing or bronchoalveolar lavage was performed on the same pulmonary segment before performing TBLB. RESULTS: Adequate specimens were obtained in 18 cases (78%). TBLB provided specific diagnosis in two cases. The results of TBLB suggested the underlying etiology in 9 cases; bacterial pneumonitis (4), hypersensitivity pneumonitis (1), polymyositis (1), radiation fibrosis (1), idiopathic pulmonary fibrosis (1), and BOOP (1). Therapeutic decisions were altered in 11 cases (47.8 %) based on the TBLB results. Pneumocystis carinii was found in the BAL fluid of another case. Ten patients with a therapeutic change and ten patients without a management change had mortality rates of 40% and 80%, respectively. The APACHE III scores were significantly higher in patients with complications (72.8+/-21.8) compared with those without complications (48.3+/-18.9) (p< 0.05). The complication rates were higher in those with mechanical ventilation (50 %) than in those without mechanical ventilation (33 %)(,) but the difference was not statistically significant (p= 0.3). Conclusions: TBLB may be a useful diagnostic option for critically ill patients with unknown cause of pulmonary infiltrates. However, it should be be used with care for patients with mechanical ventilation or for severely ill patients.